Google DeepMind has integrated "Computer Use" directly into its Gemini 3.5 Flash model, enabling it to see, reason, and act across browser, mobile, and desktop interfaces. This new capability allows developers to build custom agents for complex automation tasks like software testing and enterprise workflows. The model achieves a score of 78.4 on the OSWorld benchmark, placing it on par with GPT-5.5, and Google has implemented safety features to mitigate risks such as prompt injection. AI
IMPACT Enables more sophisticated AI agents capable of complex, multi-interface tasks, potentially accelerating enterprise automation and software development.
RANK_REASON Frontier-lab model release with new capability and benchmark.
- Browserbase
- Gemini 2.5
- Gemini 3.5 Flash
- Gemini API
- Gemini Enterprise Agent Platform
- Google DeepMind
- GPT-5.5
AI-generated summary · Google Gemini · from 8 sources. How we write summaries →